Skip to main content

tv   Sinan Aral The Hype Machine  CSPAN  December 12, 2020 11:00pm-12:31am EST

11:00 pm
's freedom and liberty and juice and equality. so ty were perfectly rea to start a newation based on those principles and that's what they did and it is a contradiction but i'm sure glad they did it.
11:01 pm
>>o the computer histy museum i hope everyone is well and fe but the pandemic we are close but the dital doors are wide open i'm pleased to welcome you today featuring the director of mit initiative on the digital economy and conversation with pulitzer prize winning journalist markoff. and those with the generosity under these unprecedented times we need your hp more than ever to sustain museum to dever on our mission for technology for everyone. in the future impact on humanit humanity's please continue to join and giv
11:02 pm
now to iroduce mr. hancock to introduce today's program and speakers. >> i am delighted you can join today's program. and deeply influenced by social media conctions sometimes a love hate relationship and the power for od. and directorf mit initiative with technology writer for questions such as e aftermath of the us election and interference in 2016 and 20 what steps can we take a look at social media companies do on the platform without cial and with these actions
11:03 pm
we can take to have the negative effects of these in her own life so for the past two decades also worked with facebook and yahoo and snapchat among other companies with a wealth of da with the new book the hype macne. and then to argue it is not inevitable. so now we'll introduce the five members all scientist and engineer investors with professor of managemenat mat. and the founding partner. and the number of startups he has founded and those that have deved to protect troops
11:04 pm
and second the ranking mber and with that complication of 18 and now it's goodo see you live and technology writer and for "the new york times" and longtime friend we look forward to your conversations. >> i really wish we were doing
11:05 pm
thisn three dimensions instead of two but we will hang in there a little bit longer so what is it like to go on a virtual book tour? >> relaxing and stressful all the same time. and then all around theorld. >> and as a data scientist and what about that she will are going from bookstore to bookstore? >> it's a little of both. it's more efficient in terms of the reach you don't have to travel than a lot of people like you just mentioned crave that more intimate connection and with ideas and to be in the presence of the body language.
11:06 pm
and i think we are missing and a lot of dimensions in our lives right now. >> but this technology has evolved more quickly so i have seen these efforts to bring the audience in. >> it's almost as if we want the auence and to fear that professionalism and i think we will navigate that in the weeks and the months and years to come. >> kevin kelly probably has the and the best of anyone that i know use the media to promote his book it has become aestseller that has been incredibly efficiently.
11:07 pm
>> i will send you his press release on this it is straightforward. i see you everywhere on social media these days you are halfway the. with the social network he began as an economist. >> i got my phd in managerial economics undermine a phd advisor i was studying to subjects at the same time and on one hand studying traditional statistics and econometrics that all the metal on - - models were known as those independently distributed so all the data independent of one another and then looking at the strategy the network of the interdependence a lot of the
11:08 pm
variance in these models that assumed by the tremendous interdependence that this is more and more true than technology so in the year 2000 and to say digital social networks that i don't know anything about that and i have been studying it ever since. >> but what about the crossover you do this interesting work to use social networks. >> yes we had a number of
11:09 pm
different collaborations over the years and we continue to collaborate. the study you are referring to was exactly that to infer networks of interaction not physical events or even in a physical space and with the long-standing work with tom allen at mat with the flow of technology and that impact of physical space and innovation and this is the modern version and those that are conferences papers on that. >> tt was very interesting work. to go back and for to be a data scientist and entrepreneur do you get them
11:10 pm
mixed up? >> that's a great question. i was the chief scientist at two startups while i was getting tenure and we were successful with a lot of work but in essence those entrepreneurial ventures flow directly so both of the startups were based in large part on scientific work tt i was doing so they go hand-in-hand so this fantastic feedback loop so the motto of mit is hand in mind and the wing - - learning by doing so the classical experience of implementing these ideas in the marketplace every day also impacts my teaching so i value
11:11 pm
the feedback loop between the marketplace and the research to keep it current and relevant keeps the teaching current and relevant and vice versa. >> do you ever worry but it is stanford but the idea they come and do the toe touch. that undergraduate class for six weeks. >> and inhose.com phases in terms of the original role of the academy? >>. >> i am a scientist entrepreneur and investor. and with a very close friend of mine the scientist
11:12 pm
entrepreneur and investor in that order. and i take it is my top priority. and that is the engine of everything else in that innovation and the toc work in and the lifeblood of everything else. and was very honored to win the award in 2018 and i take my teaching seriously. >> you did a good job but that made it stand out. what is a hype machine?
11:13 pm
>> because the business model social media is based on engagement. we are living in the attention economy in the business model of the social platform are two and then sell that attention as advertising inventory and with those that want to promote common use and then only two countries to promote geopolitical aims. and then to hype us up. and in the short term to generate the engagement which is crucial to the business model. >> obviously in the real world
11:14 pm
you cannot escape the red flag. it does feel like we are buried in red flags. and how they disrupt the elections are the economy and the health and that is a good summary that's where the danger lies. so democracy we should be concerned about interference and foreign interference and the spread of falsity if that is generating movements internally domestic terror movements to kidnap the governor of michigan planned over facebook and the momentum for that all the way through to political polarization with the algorithms on the united
11:15 pm
states and certainly the economy. so the book tells the story of the tweet put out by the ap 2013 that says barack obama was injured or killed and two explosions in the white house and the went viral. it was a real news but fake news put out by syrian hackers who hack to the ap twitter handle. and then those that were legion two other systems like trading algorithms that trade on the sentiment and they get a hold of the fake news and $140 billion of equity value and that's from a single tweet imagine we live in trillions of tweets every day small and medium and large enterprises
11:16 pm
and finally obviously there is a lot of coronavirus information and we track at mit the largest global survey and partnership with the mit additional economy in one thing we're tracking and announcing pfizer last week god 90 percent on its vaccine about other vaccines but it is related to misinformation and before that it was measles. there is a lot of misinformation and vaccination in spreading and that caused a resurgence even though we had eradicated measles in the year 2000. we saw 1800 percent increase in 2019 and a lot of that is
11:17 pm
caused by the threat of misinformation. >> are we talking metaphors or information affecting the biological world and information and virus is a real world now? >> is definitely real world now people ask me and the first time the phrase fake news was written down in the 1930s but the difference today is the speed of breath and depth to dramatically change the nature over the last ten years and essentially created a central nervous system for humanity with this technology and i think that's very
11:18 pm
different than anything we've had before. >> what is the height loop? >> we wanted to go under the hood of the tech knowledge on - - technology and so one of the important things of this machine is the algorithm is the friend recommendation for the people that you may now or the news feed. to determine the structure of the human structure network online and that and i try to simplify this with that
11:19 pm
descriptive narrative and that height loop is a dynamic interplay with the machine intelligence on one hand in human agency and how we are reacting to things on what to read and then if we take those recommendations and then that feeds back into the machine side which senses what we like and what we don't like. and to understanding the height loop helps us to understand how the height machine and then with the economy or public health. >>.
11:20 pm
>> and with the wisdom and the madness. and the fantastic book that was published in 2004 and the idea that crowds can make better decisions and to live at the truth faster. and the independence of the crowds opinion and the diversity of opinion and the quality of those individuals in the crowd. but the book was published in the same year mark zuckerberg founded facebook social media eroded all three of these principles.
11:21 pm
and what we like and abide by the confirmation bias in the two bits of science and with the relative diversity but that experimental evidence that shows the algorithms put some in the filter bubble that we narrow what we read and are exposed to those are different from the own and that is the definition of polarization. now in large-scale experiments with the hatred of the other
11:22 pm
party. but the way they read and think and then to bring it together is common ground. but then it becomes very difficult. >> a couple of years ago at uc berkeley ai professor. those could be reduced at the code and how much more complicated is it actually? >> so with the data scientist mainly which is the successor the effectiveness of the predictive power of algorithm
11:23 pm
and then to drive these what drives the success or failure not the specificity of the equation so what that means is that this such a complex set of data on which these algorithms are operating that they are complex the granular multifaceted and multidimensional that is the nature of the complexity that we see today. >> and to control this polarization the fact and the spread of this information. >> and take these in reverse order there are a number of things that we can do. so we know large-scale scientific studies have done a
11:24 pm
lot of great work that we cannot believe this information and as much and like asking people every once in a while to evaluate if a news story is true or false will get them into that reflective mode so they are less likely to believe those fake news stories and then that puts them in the reflective mood to the sharing of false news but at the same time if you do this at scale you could collect many different labels of falsity and you can see that into the machine learning algorithm how to label them and also you can use a growing number of employees as a human in the
11:25 pm
loop system. this and asking them what they are true or false and to implement that in machine learning so labeling is important to gives the providence of information and then to make better decisions. and it is extensively labeled by law. if it's produced in a facility that has wheat or peanuts if you have an allergy. but the one problem with labeling is the implied truth the fact that if you start labeling things and you don't scale it quickly is the
11:26 pm
implied truth the fact that if you start labeling things and you don't scale it quickly you have to scale to be effective in a human machine and then to do you monetize fake news you cannot have ad revenue with false news. and then in search results literacy is important and institutionalize and then i have a seven -year-old son and i'm very concerned about that and so on. that is the important first step in terms of fake news but with polarization what we found in our own experience want to turn the algorithms off so we don't become permanently change by those algorithms. so adjust the algorithms with
11:27 pm
multi- various objectives the put some more weight and importance on diversity with the alternate viewpoints to help us in terms of the polarization side from a technical perspective. >> a facebook executive told me they limit the viral spread of all kinds of things. what about that quick. >> it's important i do mention this at length in the book. the study of the false news online found falsities spread further faster deeper and more broadly than the truth in every category.
11:28 pm
and doesn't ever catch up to the falsity. and that is important way to allow the truce to catch up to falsity and then to stem the coronavirus misinformation and with the election policy changes where they made it harder to retweet in the new half to and in addition before you read tweeted it. and the articles i have written and this is to slow the information down because falsity travels farther than truth.
11:29 pm
>> is there any research does it have any impact? >> and asking you to read the reading before re- tweeting it is a good first step. >> and with that social science should shoulder for the emergence of this machine. >> all of the technologist in social scientist should and do take very seriously the responsibility steering social media of the promise to the apparel many of my colleagues spent four years about these
11:30 pm
dangers and to what extent can we create a meaningful industry academic partnership to the point where the platforms are listening and also implementing the recommendations so the academic partnership for stanford and gary king at harvard and many others to get data from facebook was effect on democracy more transparent with a lot of roadblocks and the thunder threatens to pull out of social science. and then give data that was very large amounts of data. and that partnerships to work better. and then asking them to be
11:31 pm
transparent then to be ultra- secure and private about the data. but at the same time to say lock everything down don't let my private data out. so they need to to be more transparent and more secure at the same time the way you can do that with technology or approach of the differential privacy to provide data in a way that is non- identifiable and that's how they release so this is the essential paradox
11:32 pm
because if you remember the cambridge analytica a scandal began with the university research team getting access and then passing it to cambridge analytica us. >>. >> how do you assess the political impact of the president's twitter account? what kind of impact does it hav have? >> it has completely changed political discourse. this is the primary channel through which the president of the united states communicates with the world whether it is his own staff or the public and foreign leaders, et cetera this is never been the case before. it has a tremendous amount of implications to understand and
11:33 pm
how quickly it moves in some circles but not others. and the important of the platform like censoring or laboring on - - labeling tweets of new policies. but that is important now because the presidential communication and a lot of circumstances but then they go out at 3:00 o'clock in the morning or 5:00 o'clock in the morning on a whim this is an important consortium consideration going forward for geopolitical security for domestic security and so on and these are critical still didn't discover he had a simple password on his twitter account?
11:34 pm
it was like trump 2016 or something very simple. [laughter] so what i saw in your book president trump has 89 million twitter followers but not all 89 million see every tweet. how do they decide who sees wha what? >> the algorithm behind the social media platforms rank by relevance and then tailor each feed to the individual so see the tweets in the order that is determined of an algorithm protecting what is the most relevant for you and that has a number of different features some of those might be deemed relevant and some may not. it's not a guarantee everyone
11:35 pm
sees every one of his tweets but those that they have noticed and that is true with donald trump as well. >> and big brother created a language called we speak and with the closure. but by limiting the size and expression with the nature of communication. >> and as soon as twitter came out we saw a counter idea and
11:36 pm
the idea was we will bring back longform communication in a social one - - social network style writing longer pieces and sharing them in the same way that twitter does in much shorter dialogue with the digital social network. we have an ab and flow with communication and we have a variety of tools at our disposal to communicate. and the ecosystem remains vibrant. so we don't collapse to the least common denominator and that we have multilevel communications sewer right to
11:37 pm
worry and focus on that. >> and with great interest with a meaningful impact on the election. do you come down on one side or the other? >> and the end of reality is one chapter talks about the spread of fake news and the election and the economy and health. and whether or not it has an impact and then basically there are three takeaways from this. does it vote choice or does it affect voter turnout and is that target with political campaign or misinformation and does it well targeted enough? and then the likelihood is
11:38 pm
very small most of that scientific evidence that persuades those messages that have very little effect they will switch from being a hillary clinton supporter from being a donald trump supporter whether social media or otherwise. but the results on shoulder turnout are that experimental evidence to have a meaningful significant socially significant and statistics statistically significant turnout and then it shows it can affect voter turnout dramatically and there is a significant amount of evidence and it was very broad reaching 126 million on facebook 20 million on instagram.
11:39 pm
and it was targeted at swing states and voter suppression n-95 percent of those that affect voter turnout and not pro-choice so the impact of information on social media so we don't know in conclusion what the effect was but certainly the way it would do that if it did do that is through targeted impacts on voter turnout rather than choice there are also studies in there is for the national academy of sciences that it may work for commercial advertising but it has some criticism that doesn't account for causality because they
11:40 pm
target this that is most susceptible to it. >> around 2010 first amendment research when they were persuading people to vote with this feel-good story about the propensity to vote by this wonderful social network and nobody thought at that time about voter suppression. and then to go down to the precinct level and with that facebook interaction. >> with that investigative journalism to draw out a specific narrative of that study and it involves 61 million people and that is
11:41 pm
a very large number when you consider george bush be al gore by 537 votes. >> in terms of the different social media where is the impact these days? >> these are two very different animals with a broad reach and the difference was social media is very large in scope and simultaneously tailored and targeted meaning none of us see the exact same thing on social media and all that targeting and tailoring
11:42 pm
, we don't know who sees which messages or realities people are exposed to and that's what makes it so different to taylor and target the's message is very narrowly and specifically and it's completely okay there is no transparency so we don't know who's being affected in what way. >> to that point you know if there is evidence of systematic misinformation campaign around covid information? that their campaigns run by other countries to persuade people not to wear masks in america today? >> yes there are a number of covid misinformation campaigns is not exactly clear were the origins are. that russia is much more sophisticated in 2016 but in
11:43 pm
essence and then nudging american citizens and impersonating them around those policies around that content and moving the servers to domestic soil because intelligence communities cannot look domestically they have infiltrated the cyberwar department to make it look like a missile comes from tehran. and then those misinformation campaigns that are spreading and with those effects and others that don't cure on - - that don't work and someone. >> and in 2016 with the trump election and went from silicon valley relatively quickly do
11:44 pm
no wrong to do no right that raises the question of breaking up the platform. you had an interesting business review maybe facebook is not the right approach. >> a decade of utopianism with that cool by moment when it first came out and then recently social media has been a pariah. and what my book tries to do is transcend social media a good or evil and it is yes. my book tries to answer what can we do to achieve the promise and i argue in the book that the entry ticket to
11:45 pm
solving this crisis is to create competition in the social media economy. when you say that the first thing people ask you is about facebook people think of antitrust and facebook but if you really study the economics of this market that runs on network affects the value of these platforms is the function of the number of users. so the reason they get so big there is a winner take all affect. a network effect. if you break up facebook that leading platform doesn't change the underlying economics so the next facebook like company will fall into dominance. so we have to have structural reform. what do i mean? affordability and interoperability.
11:46 pm
and that means number one we have to make the platforms more interoperable and sitting in front of congress in making any platform greater than 100 million users forced to be interoperable by law. we have done this with aol time warner merger we forced the instant messaging to be interoperable with msn and it went from 65 percent market share within three years with new startups with facebook and google we also see affordability. so you didn't use to take your number with you going from one cell phone carrier to another and we instituted portability. that was your social network at the time because all friends knew to call you at that number so that was for
11:47 pm
the cell phone age. studies show number portability created $800 million then consumer surplus every quarter for years and years after a number portability was legislated in europe. so they need to be interoperable to communicate very easily and we can take our social networks and data with us switching from one to another and this will enable consumers to choose enforce the platforms to compete like better privacy or less fake news or more security and less interference. right now we don't internalize the cost or the impact they
11:48 pm
are creating on democracy. it's like a company that pollutes into the atmosphere and makes the air less safe to inhale it is a market failure. we need some regulation. that doesn't change the underlying economic or the tendency to be less competition. >> and to questions about the mechanics of data affordability. that seems more complex i would like to take my graph with me but every one of those is somebody else so how do i take that relationship with out there okay is it simple or complex. >> it is more than taking a number with you but not only is it achievable but analogous
11:49 pm
josh calls and identity portability don't think of it as taking your network but the right to communicate with people in your network from one place to another and enabling technologies that allow that communication to happen by law. so imagine a set of protocols that would enable all of the platforms to deliver on a standard set of messaging format were there was a stack that everyone has to be able to allow the members to communicate on other platforms and then they can build on that with private proprietary messaging only available to members of their platform but
11:50 pm
that standards that would create interoperability and then you could port your identity with you from one to the other. >> is there a best way to do this? and with the antitrust action? >> they say that the case is the remedy with the antitrust lawsuit maybe the impetus to change but we need legislation for interoperability and we need legislation for identity portability and social network portability. the reason is these platforms do not have the incentive to change otherwise. the other thing i will mention because the social media economy runs on the network effect there is value created in the connections that they enable. research done at stanford
11:51 pm
estimates facebook creates $370 billion of consumer surplus every year in the united states alone. imagine that globally worldwide. and facebook is the internet and in some parts of africa facebook is the internet that is a ton of value if you tear apart those connections by breaking up facebook you will destroy value. but if you legislate interoperability and then think about antitrust that would destroy that value because those connections would be maintained because we thought the would be the tool to push into the world with a
11:52 pm
rich irony and it permitted lots of other things to come home. how do you square that paradox? >> that is a recurring theme of technology it is not deterministic good or bad so tongue in cheek good or evil the answer is yes. so what i note in my book is we have four lovers available one - - lovers - - levers available to determine the users and platforms and the advertisers and code is the design of the platform and the algorithm and with any other
11:53 pm
technology is agnostic and is a tool and then to steer it toward the promise that is essentially the theme of this book and the one sentence summary of the book. >> and that freed the platforms for responsibility. >> i go into this in detail in the book section 230 of the communication and decency act is what protects the platforms from civil liability created by users when their platform
11:54 pm
and that is essential to a free internet and a vast majority of businesses. it is the entirety of wikipedia and all of the reviews and many other businesses. section 230 is essential that is essential on the social media platforms. you can't yell fire in a crowded theater. there are places where the first amendment contradicts the 14th amendment. we have laws against defamation and libel and how do you circumscribe section 230? the inappropriate way is a
11:55 pm
five person appointed commission by the fcc because they are politically appointed commissions. a much better place to do that is the legislature and that sex trafficking law that was passed that made it illegal to deny section 230 protection for any speech that would promote sex trafficking. we need to circumscribe in ways to reduce harmful but should be done in representative bodies that dramatically politicizes it and the supreme court has a role to play to interpret the relationship between the first and 14th amendment and the limit of the first amendment as well.
11:56 pm
>> so bill gates is famous when he was under scrutiny for his monopoly that it could disappear overnight and it did. those that are very similar and i see those social network startups sitting around the edges and without getting traction but they might get traction that is any of them a threat as a significant player? >> yes. >> what we see is new social media platforms take advantage and leverage the network effect nature of the economy
11:57 pm
to rise rapidly by differentiating what type of communication and they are offering so the perfect example is tick-tock. it is short form video that none other were built to do well this is not replicated even in instagram in any way people expect certain things from it and then don't expect that from the platforms that exist there is a short form and in the long form and then twitter is about news and information and a broad scale news and information network and tiktok is the short form video content.
11:58 pm
since it is not replicated well with any other platform the network effect made it very large very quickly. so there are avenues for startups to rise but i think we need to create an environment where they can thrive so they are snuffed out in the crib as some people described by the dominant player in the way they do that social network portability data portability and interoperability. >> how do you feel about tiktok sometimes i think it dives in ways i cannot understand the something is going on. >> all of these platforms began esoteric or specific but then they expand to
11:59 pm
incorporate many more uses and that's where the norms come in how do the users adopt an appropriate and you will see that evolve over time. >> use the term hyper- socialization and it made me think what i read decades ago the social scientist at stanford that wrote about television but the internet has an either/or way he was arguing the prize meant there was less face to face interaction so where are we now? is it a bad thing that people don't spend as much time?
12:00 am
with a pandemic aside. >> so in that situation worldwide and we might have to return to that and that is dismaying but the jury is still out but it's different in different forms of interaction and communication complementing face-to-face and they give us more avenues to connect my we cannot continue to connect face-to-face and to experience enough ways that human beings will adapt and
12:01 am
balance the need with digital communication. i don't think it is necessarily good or bad it just is and we need to get the best out of it how it affects our society and cooperation and so on and we can achieve the promise. >> when you track the evolution going from desktop interaction to the palm of our hand and now they are arguing to have the computer interface. i don't know how close that is 20 or 30 years apart but i
12:02 am
worry as we find yourself closer to these machines the board comes to mind. had we keep our separate identities and technologies? >> as a huge fan of science fiction sometimes you are predicting the path that is an important question but there is a unique system of companies with the computer interface that is coming up. in large part it will begin what is known as neural
12:03 am
feedback. and then to focus on your attention or engagement or reducing stress and the easy interface is earbuds which we all wear that could feedback information and in ways that would do these things but then imagine connecting that to the always on devices like bluetooth in your home. in your home. . . . .
12:04 am
12:05 am
leverage in traction the economy and a way that serves humanity rather than the other way around. rather than humanity serving this economy. i think the true leaders of the new social age will be the ones that realize that aligning shareholder value
12:06 am
long run shareholder value with societal values is much more sustainable than aligning shareholder value with the short-term profits and engagement model that they have now. the reason for that is because of the backlash. the reason shortsighted vision is not sustainable as the delete facebook movement, the regulatory backlash the employees walking out on the whistleblowers that come from within said i can't stand this about my company. and so,. >> host: that is interesting. let's bring the audience in. i've got questions put to me in the chat channel and go to put them onto you. there's research that shows that more often a lie is repeated the easier it is for
12:07 am
the human brain to believe the lie is true because the pathways forced. do they have it obligation to prevent the spread of lies on the platform? how can we teach our children and ourselves? >> yes. this is research i know well and i can cover in the book. and yes, repetition creates a belief. now as i mentioned in the book for the last several years there's an underlying fundamental difficulty when you start to say the platform have a responsibility or should be stopping the spread of lies. but then the underlying question becomes who gets to decide what is true and what is false? let me the government should be facebook schlepping fact
12:08 am
checkers and whose checking the fact checkers? the deeply ethical important strand of the answer to that question. so cleaning philosophy and ethics at the core of this information online. if we were to agree on things that are true or false, what say they're just some things we know are false pride than i do believe the platforms have a responsibility for stopping, labeling, making it more difficult to share. make it more difficult to access those lies. because i do think faulty has the ability to dramatically affect societies outcomes and negative ways. on so yes, i do believe conditional on being able to
12:09 am
ethically arrive at what is true and what is false in a fair and equitable way than yes the problem of responsibility to any and as a society need to think very carefully about stopping the spreads of lies online. >> this question follows up on that. why do we have to accept the social media business model based on algorithms which promote outrageous inflammatory posts. why not ban these engagement algorithms instead of trying to police the content or threatening the platform companies with antitrust? sweeter that's a really, really good question. i think the answer goes to primarily that if you eliminate these algorithms, than the concept of relevance goes out the window. i knew are stuck with a reverse chronological newsfeed. that is the main problem and the genesis of the algorithm to begin with.
12:10 am
the space of information is too vast. so if we were just given a firehose of this information in reverse chronological order , it will become useless very quickly. so we have to have some way of a rank the relevance of this information. that is where you get the need for an algorithm to begin with. then the question becomes what should the algorithm look like? and there you can start to ask a much more nuanced question which is how do we write these algorithms so they are not based on hyping us up, engaging us for the short-term profit oriented business models of the economy that run on hyping us up. how do we create a multi- objective function that favor other things like more access to the university more access to truth and so in.
12:11 am
i'll give you another example about design. the like button this drives me insane. why do we have a like button? but we don't have a truth but in this top me something button or this improve my health or wellness button or whatever. and in fact the reason is because the business models are based on engagement. civilly prioritize in society today apparently his popularity. that makes my brain explode. why is that what we prioritize in society? if we had these other buttons of people be created to motivate the top people things improve their wellness and told the truth? with those be the influencers of the new social age if we prioritize the design to make those the people that one and the sort of number of kudos column? and so these are design questions that matter. just to summarize and answer
12:12 am
the question i don't think it's a good idea to ban the algorithms. because then we are back to a stream of information that is irrelevant and very difficult to get value out of. you're going to have to help me with this question are not short tracks. says okay for slightly public spaces like twitter and facebook, how about the closed groups about what apps group. i'm not sure i understand that. >> i'm not sure there's not really algorithms going on in whatsapp. so really whatsapp is a group oriented or individual oriented kind of like text messaging. there is not a relevance algorithm that is filtering and ranking content on whatsapp as it is on instagram or facebook. but there are private spaces.
12:13 am
their private accounts on instagram and so on. and algorithms are happening there as well. what is the principal difference regarding concerns against this decade in the time when the first physical height machine, the printing press of the 15th century was invented? >> the main two differences are number one the speed. the speed with which information moves from the planet is faster. therefore we run into a number of difficulties including that falsity outpaces truth. that is just one of the implications of this speed. the second is this opaque tailoring and targeting were not everyone is aware that people get different information. people don't know who is getting what information. i think the third would be the targeting it's based on these algorithms that are giving people more what they want.
12:14 am
and encouraging them to connect with people like themselves. as i describe in the book when these algorithms are based on that principle you get recommendations to connect more people like you. and then you get more information is more of what you like. that tends to reinforce what you already believe. makes it less common ground with others. so all of these implications are not necessarily true with the printing press. and so think there are major, major differences we have to struggle with. >> humans living true false information and feeding this into learning models, wouldn't the strategy be vulnerable to partisan opinions? especially when supporters are active online. for example large push on the american population claims the 2020 elections were rigged. wouldn't this drive the machine learning algorithm to an unfounded conclusion? >> yes. this is something that absolutely goes the point i made earlier. which is who gets to determine
12:15 am
what is true and what is for? how do we make sure we do not run away with the jordy opinion of people believing the earth is flat. if everyone agrees the earth is flat is that make the earth fla flat? therefore we allow that to spread? this is what i mean by very consciously and rigorously and ethically dealing with the question of how do we determine truth and falsity? i do believe once we establish something important as being true and something on the same topic being false that we need to make sure to enforce truth. the big question the open question, maybe somebody will get a nobel prize for one day, how do we equitably verifiably and scalability determine what is truth and what is false in a way that benefits society?
12:16 am
>> host: if facebook friend algorithms create tribes, how does facebook generate the hate and anger that largely drives traffic to their site? do we really expect facebook to dismantle this one has 22 billion in profits? >> i think part of it as there are echo chambers. people share this information with others who believe the same thing. sort of hypes that group up to believe a certain thing. we are seeing that happening right now with the idea of whether the election was rigged. we are seeing that information spread and echo chambers. people hearing it over and over again, repetition causes believe it causes more sharing and so on. until we see this happening. can we really expect these platforms to changes? the answer is no. i am a student of free-market
12:17 am
economics. one of the most important lessons in that study is the notion of market failure. this is a market failure. the platforms do not have an incentive to change this. we need intervention. that can be intervention along the lines of norms. people creating backlash they stop hate for profit movement, advertisers pulling out money, employees whistleblowing, leaving in droves not joining, the smartest people not joining in regulation. all of these are examples of ways society deals with market failures for this is certainly one of those instances. >> is maybe a little too soon but i'm going to ask it will talk about it. since we are also from the aftermath of the u.s. election, can you share some data run the various ways social media specifically affected the selection? such as voter turnout or voter choices? >> unfortunately too soon.
12:18 am
i think there is a number of bits of research going on. we are certainly tracking this data and so on. i think it is just way too early to tell pre-these type of studies need to be very large-scale that he should be done right the need to be meticulous. i think this is going to be one of those asked postmortems rather than real-time reporting on the effects of the election. we are just not that capable of doing a rigorous large-scale studies like that in real time. >> there enough. here's someone who wants you to put your investor hat on. what is he finding most interesting as he looks at start up companies? >> obviously i am a partner at manifest capital but i started with a longtime friend and business partner.
12:19 am
what we really are interested in is the application of ain machine learning to enterprise performance, to marketing automation. and really there we are interested in companies that have technologies that meaningfully move the needle on the ability to predict and therefore change or affect performance in terms of immeasurable kpis. we are interested in business models that are service oriented, that are subscription -based. software is a service that have recurring. and they are making meaningful changes in the performance of marketing, the performance of various enterprise endeavors. we have a bunch of investments
12:20 am
for instance in corporate gifting. we have investments in digital marketplaces. investments in marketing technology. all of them have sophisticated machine learning and ai under the hood. we think this is obviously sort of the major revolution in business of our age. there will be many, many such companies. if we can get good at evaluating them and how they are implemented in ways that create leverage for that technology into a business context, that is our strategy. >> so social media peaked in about 2011. are we anywhere near the peak of machine learning? >> no i don't think so. i think that right now what's happening is the sort of general approaches are finding applications. and that is really where the rubber is meeting the road.
12:21 am
and there you really need to be savvy thinking about how the technology that we know so well works in any particular application context and that is really where the smart and bad decisions are being made. >> here's a question from a reader actually. i'm halfway through your book great is honestly the essential reading of enjoying the thinking and the writing. my question, if you're designing a new platform today what would you do differently keeping in mind the platform would need to grow quickly and be profitable for it let's legislative changes do not happen. [laughter] thread that needle. >> that is a really good one. i think the best thing i could say about that is to think about the vast scope of human interaction and communication. and think about a mode of communication or interaction that is not well served by the current platforms.
12:22 am
the example we talked about tik tok is a great example of that. what kind of human interaction is not well served by what exists? and then think about building a platform that serves that narrow niche of interaction that people cannot get anywhere else. that would be my strategy for building tacoma mit classes. [laughter] that is good. so this might kinda follow on that. this is the last question. algorithms are used to filter huge amounts of data can algorithms be provided by customers can we flip this model in some way? >> absolutely. fantastic suggestion, one that i think is really important that i been thinking about a lot. you know, twitter has jacked dorsey said in congressional testimony and has been saying
12:23 am
for months on end now, that for instance we should be able to choose the algorithms that tailor our newsfeed among a suite of algorithms. then we could imagine that the crowd sources those algorithms so people can submit those algorithms to be part of the suite of algorithms users can choose from. i think this is a very good idea. i think we should pursue this idea. but i also think in a concept here we could add his transparency. we should be able to examine the inter- workings of these algorithms, how they process data and what they mean for the recommendations that they would make. transparent algorithms that are then chosen by the user. i think that is a fantastic idea. we should put these algorithms in a skinner box and see how
12:24 am
they behave under different conditions in order to really know what we are choosing. >> that is great. thanks for this. the make the stuff is more transparent which is getting our first step around where we are. the museum has a tradition called the one word initiative. it's going to be my job to ask you, you are asked to write down one word of advice for a young person starting out in their career. can you share your one word in the story behind why you choose it? >> absolutes flash it up on the screen? >> go for it. here's my one word. that is great. what is the story? >> the story is very simple. no young person, no person, certainly not me succeeds or fails even on their own.
12:25 am
everything that i've ever accomplished in my life has been done in teams. has been done standing on the shoulders of giants of those who came before me. has been a function of the people who taught me everything that i know. so i think that gratitude just sort of creates a sense of humility. i think gratitude is important to express because it is important to create effective teams and well working teams. and i think that gratitude is an important element of being a sort of good human and that we need to recognize all of the people around us that make it successful. and that we make successful. in that gratitude can go a long way to just in essence creating that energy and that fuel for success.
12:26 am
and also for giving back. and also for having humility. >> thanks very much for that. also, could look on the rosia virtual tool. let me bring back on the screen? are you still with us? >> sam i cannot thank you both enough, this to me as one of the most exciting conversations we have had in a long time at ch m. as we thought about our ongoing mission to the technology for people. what you presented today in this conversation as a framework that really was built upon the competing pass, right? the whole notion of interoperability and transport and people. i think about the world and i think about your language phrase the whole notion that people in the structures that we build for ourselves. and how we invoke in use these
12:27 am
tools, it is up to us. and the model that you present is straightforward. simple. grounded in both macroeconomics and social behavior. all of which are central. maybe think a little bit about a couple things in the past. one was the very beginning with desktop publishing when everyone immediately became a designer. and yet most posters look like ransom notes. recently norms emergent templates emergent people did not get so carried away. then y2k. we thought about this. enterprise software was sold with lack of interoperability. and then the bubble burst and the large enterprises held the industry accountable for in operability. frankly using less standards and portability of data as opposed to an argument over the technology stacks.
12:28 am
so things moved up a level. then you frame the whole idea around personal portability of the number. you are the unit of one you are the market with the cell phone. that is the foundation upon which all these things have been built. and we can look at the history and project forward that there are structures, social norms, regulatory levers. because the market failure will eventually be fixed. i really appreciate the thinkin thinking. and john your questioning i'm going to encourage everyone who cares about the future to take a look closely at your book and read it. so thanks so much everyone for your time and attention today. it is a great program. >> thank you everybody. >> bye-bye. ♪ ♪ book tvm cspan2 has top nonfiction books and authors every weekend and coming up this weekend sunday at
12:29 am
1:00 p.m. eastern vessels continued with kimberly hamlin from the aclu on the 19th amendment. in sports writers jessica luther and david sit on the political economic and social issues facing sports today at 6:25 p.m. eastern mit professor and tech investor with his book the height machine help social media disrupts our elections are common in our health and how we must adapt. 9:00 p.m. eastern on "after words" west virginia university writer author of love and wanted memoir choice and women and how she was denied productive choice in healthcare for children. she is interviewed by the senior vice president women's health policy director alina sal got a car. watch on book tbm cspan2 this weekend. >> here are some of the
12:30 am
current best-selling nonfiction books according to the "new york times". topping the listening first volume of his presidential memoir a promised land, former president barack obama reflects on his life and political career. i dispelled by actor matthew mcconaughey's memoir and green lights. after that is dolly parton song tell her what she looks to her life they were songs. also on the list of former first lady michelle obama's memoir becoming. and wrapping up our look at some of the best sewing books according to the "new york times" is actor michael j fox his reflection on living with parkinson's disease. in no time like the present. : : :

27 Views

info Stream Only

Uploaded by TV Archive on